Fix -n (Resolve #3)

README.md

@@ -21,13 +21,12 @@ poetry install --only main
 poetry run pip install -U torch torchaudio --extra-index-url https://download.pytorch.org/whl/cu116
 
 # Run in English
-poetry run whisper_streaming --language en --model base -n 20
+poetry run whisper_streaming --language en --model base
 ```
 
 - ``--help`` shows full options
 - ``--language`` sets the language to transcribe. The list of languages are shown with ``poetry run whisper_streaming -h``
 - ``-t`` sets temperatures to decode. You can set several like (``-t 0.0 -t 0.1 -t 0.5``), but too many temperatures exhaust decoding time
-- ``-n`` sets interval of parsing. Larger values can improve accuracy but consume more memory.
 - ``--debug`` outputs logs for debug
 
 ## Tips

whisper_streaming/cli.py

@@ -81,8 +81,8 @@ def get_opts() -> argparse.Namespace:
         "--num_block",
         "-n",
         type=int,
-        default=20,
-        help="Number of operation unit. Larger values can improve accuracy but consume more memory.",
+        default=160,
+        help="Number of operation unit",
     )
     parser.add_argument(
         "--temperature",

whisper_streaming/transcriber.py

@@ -239,16 +239,17 @@ class WhisperStreamingTranscriber:
             self.buffer_mel = None
 
         seek: int = 0
-        rest_start: Optional[int] = None
         while seek < mel.shape[-1]:
             segment = (
                 pad_or_trim(mel[:, :, seek:], N_FRAMES)
                 .to(self.model.device)  # type: ignore
                 .to(self.dtype)
             )
+            if segment.shape[-1] > mel.shape[-1]:
+                logger.warning("Padding is not expected while speaking")
 
             logger.debug(
-                f"seek={seek}, timestamp={self.timestamp}, rest_start={rest_start},"
+                f"seek={seek}, timestamp={self.timestamp}"
                 f"mel.shape: {mel.shape}, segment.shape: {segment.shape}"
             )
             results = self._decode_with_fallback(
@@ -266,7 +267,6 @@ class WhisperStreamingTranscriber:
                     and result.avg_logprob > self.config.logprob_threshold
                 ):
                     seek += segment.shape[-1]
-                    rest_start = None
                     logger.debug(
                         f"Skip: {segment.shape[-1]}, new seek={seek}, mel.shape: {mel.shape}"
                     )
@@ -283,15 +283,14 @@ class WhisperStreamingTranscriber:
                     yield v
             if last_timestamp_position is None:
                 seek += segment.shape[-1]
-                rest_start = None
             else:
                 seek += last_timestamp_position * self.input_stride
-                rest_start = seek
             logger.debug(f"new seek={seek}, mel.shape: {mel.shape}")
 
-        logger.debug(f"Last rest_start={rest_start}, mel.shape: {mel.shape}")
-        if rest_start is None:
-            return
+            if mel.shape[-1] < N_FRAMES:
+                break
 
-        self.buffer_mel = mel[:, :, rest_start:]
+        if mel.shape[-1] - seek < 0:
+            return
+        self.buffer_mel = mel[:, :, seek:]
         del mel